Every commit scored for change-risk against this repo's own history, so 'elevated' means elevated here rather than on some global curve.
Needs review
82 commits sit in this repo's top risk tercile, which is 34% of the 240scored. The cut is drawn against this codebase's own history rather than a global curve, so a quiet repo still fills its top band, and here it starts at 3.8 out of 10. What pushes a commit up is size and spread together: a large change confined to one area scores below a smaller one scattered across a dozen files.
Commit categories over time, read off the subject line. Fixes carry the accent because that is the series this chart exists to show.
Consistently other-driven across its history.
Ranked by change-risk, highest first. Priority is a tercile of this repo's own distribution, so a quiet repo still fills its top band.
| # | Commit | Author | When | Lines | Risk | Top driver |
|---|---|---|---|---|---|---|
| 1 | 3d59aad1adding test for latest transformer and torch versions | Gavin Li | 2y ago | +23.2K -0 | 99%Elevated | large diff (many lines added) |
| 2 | 8a4a7705add example notebooks | Yu Li | 2y ago | +47.1K -0 | 99%Elevated | large diff (many lines added) |
| 3 | c42f5ebc使用 Colaboratory 创建 | Gavin Li | 2y ago | +47.1K -6 | 99%Elevated | large diff (many lines added) |
| 4 | abfb9c0btest sealllm | Gavin Li | 2y ago | +4.0K -0 | 98%Elevated | large diff (many lines added) |
| 5 | 21c01a3dadding compression test to make sure prefetching doesn't break compression | Gavin Li | 2y ago | +5.6K -0 | 98%Elevated | large diff (many lines added) |
| 6 | 834d5e7acs | Yu Li | 3y ago | +5.7K -0 | 98%Elevated | large diff (many lines added) |
| 7 | 21916bc0update infer notebook | Gavin Li | 3y ago | +6.7K -105 | 98%Elevated | large diff (many lines added) |
| 8 | db7a1d13add eval dataset, eval code, elo rating code | Yu Li | 3y ago | +13.5K -1 | 98%Elevated | large diff (many lines added) |
| 9 | 44ef329dadd macos support | Yu Li | 2y ago | +4.9K -44 | 96%Elevated | large diff (many lines added) |
| 10 | cabf98f1support baichuan mistral internlm | Yu Li | 2y ago | +1.5K -9 | 96%Elevated | large diff (many lines added) |
| 11 | 64042f05init commit for anima 100k | Yu Li | 2y ago | +2.7K -229 | 96%Elevated | large diff (many lines added) |
| 12 | 41272f54auto model | Yu Li | 2y ago | +720 -2.4K | 95%Elevated | large diff (many lines added) |
| 13 | eceaaf52gen eval dataset | Yu Li | 2y ago | +1.1K -0 | 95%Elevated | large diff (many lines added) |
| 14 | 2597f17cadd infer ipynb code | Gavin Li | 3y ago | +604 -0 | 94%Elevated | large diff (many lines added) |
| 15 | b3ca4eb5code open source, update README | Yu Li | 3y ago | +990 -0 | 94%Elevated | large diff (many lines added) |
| 16 | d2182a1badd Mixtral | Yu Li | 2y ago | +741 -10 | 93%Elevated | large diff (many lines added) |
| 17 | 662d1d6eadd airllm | Yu Li | 2y ago | +701 -0 | 93%Elevated | large diff (many lines added) |
| 18 | 2b3fa3cdinit dpo based rlhf | Yu Li | 3y ago | +969 -4 | 93%Elevated | large diff (many lines added) |
| 19 | 64f3c1cdadd support of chatglm | Yu Li | 2y ago | +473 -31 | 92%Elevated | large diff (many lines added) |
| 20 | 916efd5esupport qwen | Yu Li | 2y ago | +463 -7 | 92%Elevated | large diff (many lines added) |
| 21 | 2b5d50b9add safetensors model support, fix profiling, refactor code, fix typo | Yu Li | 2y ago | +360 -268 | 92%Elevated | large diff (many lines added) |
| 22 | 6d65cffbUpdate README.md | Gavin Li | 2y ago | +311 -76 | 91%Elevated | large diff (many lines added) |
| 23 | 26a0a134en version | Yu Li | 3y ago | +208 -0 | 91%Elevated | large diff (many lines added) |
| 24 | 0e24882cInitial commit | Gavin Li | 3y ago | +203 -0 | 90%Elevated | large diff (many lines added) |
| 25 | 2752a41bdocs: add Japanese README | Ikko Eltociear Ashimine | 2y ago | +230 -0 | 90%Elevated | large diff (many lines added) |
| 26 | c2d13c10profiling and compression | Yu Li | 2y ago | +193 -8 | 90%Elevated | large diff (many lines added) |
| 27 | 4159a535add prefetching | Yu Li | 2y ago | +132 -48 | 89%Elevated | large diff (many lines added) |
| 28 | e222a652readme | Yu Li | 3y ago | +88 -1 | 88%Elevated | large diff (many lines added) |
| 29 | 351a1a38add instructions for finetune based on merged models | Yu Li | 3y ago | +91 -18 | 88%Elevated | large diff (many lines added) |
| 30 | aa2a6f69Add funding.json for open source funding transparency | Yu Li | 4mo ago | +74 -0 | 87%Elevated | large diff (many lines added) |
| 31 | fcef20eafix automodel on macos | Yu Li | 2y ago | +89 -1.4K | 87%Elevated | large diff (many lines added) |
| 32 | e8788315init readme | Yu Li | 3y ago | +64 -1 | 87%Elevated | large diff (many lines added) |
| 33 | 922e1c09initial add auto model | Yu Li | 2y ago | +66 -0 | 86%Elevated | large diff (many lines added) |
| 34 | 04414670llama3.1 405b example | Gavin Li | 2y ago | +68 -0 | 86%Elevated | large diff (many lines added) |
| 35 | c8a87f9asupport delete_original model files | Yu Li | 2y ago | +66 -21 | 85%Elevated | large diff (many lines added) |
| 36 | 233bed9fsupport specify the layer shard saving path | Yu Li | 2y ago | +83 -32 | 85%Elevated | large diff (many lines added) |
| 37 | 1a710a57update README | Yu Li | 3y ago | +51 -6 | 85%Elevated | large diff (many lines added) |
| 38 | 9ace0fc5support Llama3.1 405B | Yu Li | 2y ago | +51 -15 | 84%Elevated | large diff (many lines added) |
| 39 | 471a83ddadd more download files validation | Yu Li | 2y ago | +59 -14 | 84%Elevated | large diff (many lines added) |
| 40 | 57c2b6eeUpdate README.md | Gavin Li | 2y ago | +51 -0 | 84%Elevated | large diff (many lines added) |
| 41 | 66f33b67update README adding troubleshooting | Yu Li | 3y ago | +51 -48 | 83%Elevated | large diff (many lines added) |
| 42 | e4f918d8update README adding troubleshooting | Yu Li | 3y ago | +49 -0 | 83%Elevated | large diff (many lines added) |
| 43 | e2cf73d3update readme | Yu Li | 2y ago | +46 -21 | 82%Elevated | large diff (many lines added) |
| 44 | 46f17008update readme | Yu Li | 3y ago | +43 -39 | 82%Elevated | large diff (many lines added) |
| 45 | f2522ed6added support for models that are single files and not sharded in huggingface repo | Navodplayer1 | 1y ago | +32 -17 | 81%Elevated | large diff (many lines added) |
| 46 | 97e96d8bfix new_seq | Yu Li | 2y ago | +39 -17 | 81%Elevated | large diff (many lines added) |
| 47 | e14115b5update readme | Yu Li | 3y ago | +35 -35 | 81%Elevated | large diff (many lines added) |
| 48 | f738a1eecreated from Colaboratory | Gavin Li | 2y ago | +39 -0 | 80%Elevated | large diff (many lines added) |
| 49 | 693f01c6support hf token | Yu Li | 2y ago | +28 -8 | 80%Elevated | large diff (many lines added) |
| 50 | 19283f57publish 2.0.0 | Yu Li | 2y ago | +40 -3 | 80%Elevated | large diff (many lines added) |
Two views of the same model: where the cuts fall, and what commit shape lands you above them.
Every scored commit, binned on the raw 0 to 10 score rather than the percentile. Percentile ranks are uniform by construction, so that axis has no shape to draw. The dashed lines are the tercile cuts behind each row's priority pill.
The 200 most recent commits, on their own recency sample rather than the feed above: that defaults to risk-sorted, so reusing it would plot only the top tercile and call it the spread. Big and scattered is what the model penalises. Click a dot to open it.
Repowise tracks change history across 34 files in 2233admin/airllm. No file in the repository changed in the last 90 days. Every commit is scored for change risk from its size, spread and the history of the files it touches.
